Meta shares fell 11% on Wednesday after quarterly results laid bare the cost of the company’s AI ambitions: free cash flow collapsed to $784 million, the lowest in at least five years, while spending guidance climbed to $145 billion for the year.
Revenue grew 28% to $61 billion for the quarter from April to June. Profits fell 14% to $6 billion. The gap between what Meta earned and what it kept is the story — almost every dollar of operating cash got eaten by AI infrastructure.
What is free cash flow? It’s the money a company has left after paying all its operating costs and capital expenses. It’s what’s available for shareholders, buybacks, dividends, or weathering a downturn. When it drops to near-zero, investors notice.
🔍 THE BOTTOM LINE
Meta is spending like a company that has a working AI product line. It does not yet have one. Zuckerberg told analysts the company would start selling AI tools to other businesses, but the revenue from that business is zero today. Meanwhile, Microsoft — spending even more in absolute terms — showed that AI spending can satisfy investors when it comes with returns. The market is starting to distinguish between AI investment and AI speculation.
What the Numbers Say
Meta’s spending guide moved from $125 billion to $130–145 billion in three months. CFO Susan Li told analysts that selling AI to other companies would help drive returns, but offered no timeline beyond “by 2028, we’ll have turned over a lot of cards.”
The free cash flow figure — $784 million — is the number that spooked Wall Street. For context, Meta generated $14 billion in free cash flow in the same quarter a year ago. The decline is not gradual. It’s a cliff.
Google reported its own lowest-ever cash position the previous week, sending its stock tumbling too. The two companies are burning cash on AI infrastructure at a rate that makes the metaverse look modest.
The Metaverse Echo
For Forrester analyst Mike Proulx, the parallel is obvious. “There’s a bit of similarity to Meta’s metaverse missteps in that Meta is once again spending ahead of proven product demand,” he said, referring to the tens of billions Meta poured into virtual reality experiences that never gained traction.
The difference: the metaverse was a speculative bet on a future platform. AI is a proven technology with real revenue — just not yet for Meta. The company’s AI tools are boosting engagement on Instagram and Facebook, and Zuckerberg pointed to improved ad performance for small businesses. But those gains are incremental, not transformative. They don’t justify $145 billion.
What Zuckerberg Promised Next
Zuckerberg outlined two revenue paths during the earnings call. First: AI agents — chatbots that act autonomously on users’ behalf. “Soon, we’ll have agents that can work 24/7 on your behalf,” he said. This builds on his earlier push toward personalised AI bots, though his admission last month that AI agent development hadn’t progressed as expected still hangs over the pitch.
Second: selling AI to other companies. Meta plans to make its Muse Spark model available via API, along with coding and productivity tools. “We expect to build a large business for large businesses,” Zuckerberg said. This would be Meta’s first serious enterprise software play — a market where Microsoft, Google, and Amazon already have years of head start.
The Microsoft Contrast
Microsoft reported on the same day and got the opposite reaction. Shares rose 5% in after-hours trading. Sales were up 18% to $90 billion. Profits up 31% to $36 billion.
Microsoft’s capital spending guidance for next year is $175 billion — more than Meta’s. But Microsoft is showing returns: Azure AI revenue is growing, Copilot is embedded in enterprise contracts, and the OpenAI partnership is producing measurable product integration.
CEO Satya Nadella used the call to address OpenAI’s recent model behaviour issues, saying “you can’t depend on any one model” — a tacit acknowledgement that the AI infrastructure race is not just about spending, but about reliability.
The market read: Microsoft is converting AI spend into revenue. Meta is converting AI spend into promises.
NZ Angle
For New Zealand businesses watching the AI infrastructure race from the sidelines, Meta’s earnings underscore a structural reality: the companies building foundational AI models are spending at a scale that locks out competition. When Meta’s free cash flow drops 95% year-on-year and the stock still trades at a premium, it means investors are pricing in a future where only a handful of companies control AI infrastructure.
NZ companies using Meta’s AI tools — or considering them when the enterprise API launches — should note that the pricing power sits entirely with the provider. When the company burning $145 billion a year starts selling you AI, the margin is on their side.
❓ FAQ
Why did Meta shares fall if revenue grew 28%? Because investors are pricing in the spending trajectory. Revenue growth of 28% is strong, but profits fell 14% and free cash flow hit a five-year low. The market is questioning whether the spending will produce returns before the cash runs out.
What is Meta actually spending $145 billion on? Mostly AI infrastructure — data centres, GPU clusters, and the energy to run them. Meta is building capacity to train and serve large language models at scale, both for its own products and for the enterprise AI business it plans to launch.
How is this different from the metaverse spending? The metaverse was a bet on a platform that didn’t exist. AI is a proven technology with existing revenue streams — just not at Meta. The risk is not that AI won’t work; it’s that Meta’s specific AI products may not generate enough revenue to justify the spending level.
What does the EU’s designation of ChatGPT mean for Meta? Separately, the EU is moving to classify ChatGPT and Roblox under its strictest platform rules. Meta’s platforms are already designated as Very Large Online Platforms under the DSA. Any expansion of AI-specific regulation under the DSA framework could affect Meta’s AI agent ambitions in Europe.
🔍 THE BOTTOM LINE
Meta’s quarter is a snapshot of the AI infrastructure dilemma: spend now, prove later. The company has the resources to sustain this for years, but each quarter of declining cash flow narrows the margin for error. Microsoft showed the same day that the market will reward AI spending — but only when the revenue shows up in the same report. Zuckerberg’s “by 2028” timeline may not be fast enough for investors watching the cash burn in real time.
